Item-level analysis allows for the examination of qualitative age and individual differences in skill acquisition, which are obscured when aggregating data across items. In the present study, item-level strategy shifts were generally gradual and variable, rather than abrupt and collective. Strategy shift reversions were frequent, and the total transition space was extensive, for both younger and older adults. Shift indices were highly variable between items for both younger and older adults. Age differences in item-level shift patterns suggest that older adults' greater conservatism in strategy selection leads to more gradual strategy shift transitions for individual items as well as to more collective strategy shifts.
Age differences in skill and information acquisition are documented across various cognitive domains. Older adults typically learn more slowly than younger adults, and reach lower levels of asymptotic performance (see, e.g., Bosman & Charness, 1996; Charness, 1981; Hoyer, Cerella, & Onyper, 2003; Siegler& Lemaire, 1997; Strayer & Kramer, 1994; Touron & Hertzog, 2004a, 2004b; Touron, Hoyer, & Cerella, 2001, 2004). Several theoretical approaches have been advanced to account for these performance deficits, including differences in strategy choice and efficiency (Strayer & Kramer, 1994; Touron & Hertzog, 2004a, 2004b), associative learning deficits (Hoyer et al., 2003; Touron et al., 2001, 2004), and the influence of age-related slowing on component task process (Salthouse, 1994; Touron et al., 2004).
Age differences are particularly pronounced for skill acquisition tasks involving a transition in response method-referred to here as a strategy shift-from performing the task based on an algorithm to retrieving solutions directly from memory (see, e.g., Logan, 1988; Rickard, 1997). Several studies have shown that older adults' strategy shift is slower and less complete than strategy shift by younger adults. This is true for verification and production tasks involving a shift from a computation algorithm to memory retrieval (Touron et al., 2001, 2004), for tasks involving a shift from a simpler visual search algorithm to memory retrieval (Rogers & Gilbert, 1997; Rogers, Hertzog, & Fisk, 2000; Touron & Hertzog, 2004a, 2004b), and for tasks that involve the development of automaticity in memory search tasks (Fisk & Rogers, 1991; Strayer & Kramer, 1994).
Previous investigations concerning age differences in strategic skill acquisition have focused exclusively on overall task performance. However, an item-level approach could expose further qualitative differences in task approach by older and younger adults that are obscured when analyzing aggregate data. Although such strategies cannot completely account for cognitive performance (Light, 1996; Salthouse, 1991), strategic behavior has a marked impact on performance differences in many complex cognitive tasks (see, e.g., Dunlosky & Hertzog, 2001; Lemaire & Arnaud, 2002). Furthermore, aggregation of acquisition data from individual participants into group functions can produce artifactual biases and obscure meaningful regularities (see Estes & Maddox, 2005; Myung, Kirn, & Pitt, 2000), and analogous consequences may result when data are aggregated across items.
The present study examines age differences in itemlevel strategy shifts using the noun pair (NP) lookup task. Participants were asked to verify whether or not a target noun pair appeared in a lookup table (Ackerman & Woltz, 1994). Noun pairs were consistently mapped (CM; Shiffiin & Schneider, 1977) to allow learning of the associative pairings. Discrimination of matched from unmatched pairs could be achieved by either visual search of the lookup table (referred to here as scanning) or, with repetition, by memory retrieval. Because retrieval is typically much faster, strategy shift can be inferred when CM noun pair response times (RTs) become reliably shorter than variably mapped (VM) noun pair RTs (Ackerman & Woltz, 1994; Rogers & Gilbert, 1997; Rogers et al. …